The early mnemonic medium offers readers little agency

Readers of the Mnemonic medium (as deployed on Quantum Country and initial versions of Orbit) are put on quite a rigid track:

  • they’re expected to interact with all questions:
    • you can’t skip a question in the middle of a set (only the entire set)
    • you can’t delete or disable a question once it’s “collected”
    • the progress mechanics assume you’re trying to collect all the questions and “penalize” you if you don’t
  • questions are not modifiable
    • readers can’t add their own questions or situate them relative to their own thinking

These factors…

Forward directions

I’d like to swing the vibe aggressively more in the reader-centric direction. This will be tough to do while maintaining other important properties of the medium. Expert-authored prompts really do contribute a great deal of value (The mnemonic medium supplies expert-authored prompts). We want to reap that value (or perhaps even enhance it!) while also providing fluidity, malleability.

There are some relatively simple ways we might begin: readers should be able to disable or skip prompts that don’t resonate; readers should be able to make simple edits to authors’ wording; the embedded interface shouldn’t discourage people from reviewing only part of a page. All this will generate interesting signal for authors: The mnemonic medium can generate interesting author analytics

But I think we’ll have to go deeper than this. A few directions I find promising, which of course need to be fleshed out more thoroughly:

  • You should be able to fluidly move between consuming a prompt, to editing its text, to refactoring it (one-to-many) and its adjacent author-provided prompts (many-to-one), to writing new prompts of your own inspired by those prompts.
    • Ideally, this same interaction model should extend to the review experience: as you’re reviewing, if you think of a way to improve a prompt, or you’re reminded of a new prompt you want to write, that should be just as fluid.
  • We should explore pick-and-choose modalities for ingesting prompts from the text. Imagine that as you read, some phrases are highlighted in a special way indicating that the author has provided prompts representing that idea. If you find something interesting, you can click on the highlighted text—and boom, now you’ll remember it forever.
    • We’d want the same fluid consuming-to-editing interactions for this type of prompt.
    • This type of interaction points the way to crowdsourced annotations, ML-generated prompts, etc. More generally: imagine if you could point at any idea you found interesting, anywhere, and feel assured that now you’ll remember it (and ideally deeply internalize it) for the long term.
    • Ozzie Kirkby suggests that there may be some interesting middle ground here which would encourage readers to write their own prompts representing the details they find interesting: maybe authors (or ML, or the system) can provide partial/“starter” templates for key text regions, which the user would adapt/complete as they see fit.
  • We can start to experiment with fluid interaction between the medium and the rest of the reader’s knowledge work by fleshing out My implementation of a personal mnemonic medium. That direction is really not quite the same as what the rest of this note is talking about: I’m not sure how author-provided prompts interact with this kind of note-taking. So maybe it’s a red herring… but my instinct is that it’ll be instructive, and that all this work is intertwined.
  • I suspect this work will inform and intersect with the design of a “library view” in the Orbit app itself. One related priority for me here is to fuzz the “boundaries” between prompts; see discussion in “How to write good prompts” here, grep (“revision is a holistic endeavor”). This also means fuzzing the boundaries between the author’s prompts and the user’s prompts. We should probably avoid paradigms which put those sets in separate conceptual “places”—but I expect we’ll still want some affordances showing provenance.

::TODO much of this section should be fleshed out in more elaborate separate notes::

References

See notes from Yusuf Ahmad on this tension, which he frames as a gulf between “industrial” educational architectures and social constructivist architectures: https://www.notion.so/Placing-Cog-Sci-ideas-within-a-broader-Learning-Architecture-fa69fc15ba9e41b8a2e136f7fbe0362b